ThermalTake Hardcano II - goofy sensors?

[MiniMizer]

Yesterday I got myself one of those nifty things, and I installed it right away.
The first thing it did was to burn my cpu, somehow the fact that I ran it underneath the cpu caused an electrical shortage, and bye byte cpu.
It was an Athlon 1.33GHz (1.50 as a matter of fact), rest his soul and it took one my my DDR modules with it...
Anyway, I replaced the cpu today and noticed the HardCano II review on hardCOREware suggesting to put the sensor on top of the cpu.
I put it on top as suggested, used some tape to hold it in place so only it's edge is touching the cpu core.
The good news are that everything works
The bad news are that I've no idea if it works the way it should...
I have a sensor right underneath the cpu, coming from the mobo.
I use MBM to measure temperature from it.
Right now, MBM shows 55c.
Right now, Hardcano shows 47c.
I've no idea who gives a better reading, though I can tell you the heatsink (AX7) is pretty hot, and if I'm gonna take Hardcano's word, it goes about 39c.
Now, I need some advices...
Which sensor do I believe? The onboard one that touches the CPU's bottom or the one that touches it's core?
And why is there such a big difference between the two?

Edit: I'm not sure if this is the correct forum for my question, but it seemed to me to be the correct one, still, move it freely if you like.

 

[KennyH]

I am not sure, I was thinking about picking up a Hardcano 5 myself.

 

[MiniMizer]

I think you should go for it, I'm very pleased with mine because based on preliminary testing I've done (open case lying on it's side) the little fan helps reduce the temp by 4c (39c > 35c).
However, the question about the sensors remains yet unanwsered

 

[MiniMizer]

*bump*

 

[Mikewarrior2]

what motherboadr are you using?



Mike

 

[MiniMizer]

a Gigabyte one (7VTX).
I have indeed heard that Gigabyte's sensors aren't too accurate, but nevertheless, such a difference in readings is suspicious.

 

[Mikewarrior2]

Careful about listening to people who say that "high reading" motherboards aren't too accurate. Higher reading motherboads are almost always closer to reporting true DIE temperature than low-reading motherboards.

What is your Case temp? Remember that your CPU is a real hot bastard . Chances are, since Side-mounted thermistors have many variables involved, that the MBM reading is probably closer to die temp than the Hardcano II sensor.



mike

 

[MiniMizer]

My case is currently open so the ambient temp is pretty close to the room temp, which is just above 26c.
Right now, the cpu temp is either 45c (HardCano) or 51c (MBM).
The temperature of the heatsink itself is 38c (HardCano), if it helps.
What is the average difference between the heatsink temperature (measured around the middle of it) and the temperature of the core itself?
Anyway, you are correct that heat dissipates in all directions, however, don't you think that a thermistor that it touching the core directly is giving more reliable results than a thermistor which is located beneath the entire cpu?

 

[Mikewarrior2]

In general, I think Socket-thermistor's are teh worst temp reading devices ever . Die backside readings die.

At the same time that side-thermistors do measure temp change much better, the backside reading is much more inline with what your "theoretical" cpu temp is, so that's whY i said the mbm reading was probably close to die temp.

The correlation of heatsink temp to CPu temp varies greatly, but a poor heatsink will exhibit a small delta between DIE temp and Heatsink temp, while a good one has a large delta.

(theoretical temp calculation):

80W cpu at 1.5ghz, 1.75V vcore (this goes up if your vcore is higher). .35C/W for the heatsink (probably a generous value).

26C ambient case temp

~54C or so CPU temp.



Mike

 

[MiniMizer]

Ok, so suppose the onboard thermistor gives a better reading, still, what causes such a large gap in the temperature readings?
I would understand if it was 1c-2c but 8c?
Something has got to be wrong here and in my opinion, it's the onboard sensor.
I believe so because the hardcano thermistor is touching the core directly since I've removed part of the plastic around it to get an even better reading.
Furthermore, as far as I am concerned, the Hardcano sensors are reliable; This morning, after my pc was off all night, both sensors showed the exact same temperature (0.2c difference to be honest), while one was connected to the core and the other one to the heatsink itself.
I've tested them several times and in all times they proved to be reliable.
So what on earth could be causing such a drastic difference in temperature readings?

 

[Mikewarrior2]

Well, backside readings are inaccurate in terms of using it for comparison purposes. They are also generally inaccurate. But sometimes, the compensation a manufacturer uses makes the socket-thermistor a "good-guess" reading.

What side the hardcano thermistor is touching the core makes a huge difference. For example, if it is in contact with teh L2-Cache side of the t-bird core, then its reading will be lower than if it were on the side with the ALU portion of the CPU.

Secondly, no matter how good you place a side-mount thermistor, the readings will always be a combination side-core, radiant heat from the heatsink, CPU pcb heat, etc. Under best conditions (mounted on hottest side, etc), a side-mount thermistor will only show about 75-80% of the CPU temp rise over ambient.

Hence, in your case, the side-thermistor is showing 19C rise over ambient. if it is showing 75% rise over ambient, your die temp based on this reading could be 25C over ambient. A lot depends on where the side-thermistor is mounted.

But I generally wouldn't worry about the discretion in temps between the two readings. They are from different portions of the CPU (which is why external readings are faulty on all CPUs). Chalk up the MBM reading to being a "lucky guess from the motherboard compensation" and chalk the Hardcano reading to simply matching the general nature of side-core readings.


Mike